name	ry-rules-review
description	Аудит против rldyour rules. Используй для: rules review, проверь правила, compliance. EN: policy audit, hard rules check.

ry-rules-review

Review a diff, PR, branch, file scope, or implementation against rldyour-rules.

Determine target: current diff, branch vs main, PR, file scope, or user-provided scope.
Use Serena-first code inspection for affected symbols and integration paths.
Apply quality-first-engineering, architecture-boundaries, implementation-discipline, dependency-compatibility-policy, verification-quality-gates, and project-instructions-policy.
Use rldyour-explore when the review depends on current technology behavior, dependency versions, or architecture best practices.
Report findings in Russian, ordered by severity and confidence.
Default mode is report-only. Modify files only if the user explicitly asks to fix findings.

Severity: critical, high, medium, low.
Confidence: 0-100.
Location: exact file and line when possible.
Rule: which rldyour rule is violated.
Evidence: concrete code, config, test, or docs evidence.
Impact: what becomes incorrect, fragile, unscalable, insecure, or harder to maintain.
Fix: actionable correction.
Disposition: must-fix, should-fix, ask-user, or defer.

Read references/rules-policy.md and only load other references when they match the reviewed issue.

name	ry-rules-review
description	Аудит против rldyour rules. Используй для: rules review, проверь правила, compliance. EN: policy audit, hard rules check.

Review a diff, PR, branch, file scope, or implementation against rldyour-rules.

Determine target: current diff, branch vs main, PR, file scope, or user-provided scope.
Use Serena-first code inspection for affected symbols and integration paths.
Apply quality-first-engineering, architecture-boundaries, implementation-discipline, dependency-compatibility-policy, verification-quality-gates, and project-instructions-policy.
Use rldyour-explore when the review depends on current technology behavior, dependency versions, or architecture best practices.
Report findings in Russian, ordered by severity and confidence.
Default mode is report-only. Modify files only if the user explicitly asks to fix findings.

Severity: critical, high, medium, low.
Confidence: 0-100.
Location: exact file and line when possible.
Rule: which rldyour rule is violated.
Evidence: concrete code, config, test, or docs evidence.
Impact: what becomes incorrect, fragile, unscalable, insecure, or harder to maintain.
Fix: actionable correction.
Disposition: must-fix, should-fix, ask-user, or defer.

Read references/rules-policy.md and only load other references when they match the reviewed issue.

Plus depuis ce dépôt